Skip to main content

Table 3 Overall results of the proposed methods (in decreasing accuracy order)

From: A procedure to locate the eyelid position in noisy videokeratoscopic images

Rank

Filtering

Model

Fitting

Median ± MAD

1

GDV

Parabola

Hough

3.50 ± 1.37

2

GDV

Rice

LAD

3.52 ± 1.04

3

GDV

Parabola

Bisquare

3.54 ± 0.91

4

GDV

Fourth-order

LAD

3.61 ± 1.80

5

GDV

Rice

Bisquare

3.62 ± 1.37

6

GDV

Parabola

LAD

3.66 ± 1.17

7

GDV

Rational

Bisquare

3.82 ± 0.61

8

GDV

Fourier

Bisquare

3.83 ± 1.57

9

GDV

Fourth-order

Bisquare

3.91 ± 2.28

10

GDV

Skew-normal

Bisquare

4.03 ± 2.22

11

GDV

Cubic

LAD

4.04 ± 1.74

32

Wavelet

Weibull

Bisquare

4.62 ± 2.11

40

Wavelet

Fréchet

Bisquare

4.86 ± 2.27

42

GDV

Dagum

LAD

4.90 ± 1.08

60

GDV

Log-logistic

LAD

5.16 ± 2.40

113

GDV

Gamma

LAD

5.71 ± 3.03

  1. The results are the median and MAD of the RMS deviations over all ten images